Classification of HIV data by constructing a social network with frequent itemsets

Acquired immune deficiency syndrome (AIDS) is the last and the most life-threatening phase of Human Immunodeficiency Virus (HIV) disease. HIV attacks and heavily affects the immune system of the body which remains unable to resist the disease. HIV uses white blood cells to replicate itself and spreads everywhere in the body. The lifecycle of HIV disease, especially the replication stage must be prominently understood in order to develop effective drugs for treatment. HIV-1 protease enzyme is in charge of cleaving an amino acid octamer into peptides which are used to create proteins by virus. It should be scrutinized properly since it is a potential target to tightly bind drugs to protease for blocking the virus action at an early stage before cell infection. It is very critical to induce a model and predict cleavage of HIV-1 protease on octamers. Several machine learning approaches have been applied for predicting and profiling cleavage rules. However, we propose a novel general approach that can also be applied on different domains. It basically utilizes social network analysis and data mining techniques for classification. This method yet presents promising results that are comparable with existing machine learning methods, besides it gives the opportunity to validate the results obtained by using other techniques from social network analysis perspective. We have used the HIV-1 protease cleavage data set from UCI machine learning repository and demonstrated the effectiveness of our proposed method by comparing it with decision tree, Naive-Bayes and k-nearest neighbor methods.